Human agency in the era of artificial agents

Throughout my youth my parents insisted I was well educated. Neither of them came from a wealthy family and they believed this was the best chance for my future. For them, it was important that me and my brothers did well in school because, for their generation, that meant good jobs. In spite of all that, neither of them would have dreamt that I should learn how to read in order to have a job. Reading for my family has always been a value in an of itself. Neither of my parents would have argued that the main reason for me to go to school was financial health. In fact, once, when I was small and feeling that my efforts to learn maths were futile, I asked my mother why I had to learn all that. Her answer left me kind of speechless: "because you have the right to know all that." And she went on about how much humanity had learned and discovered throughout the centuries and millennia, and that I had the privilege to learn it all. She never mentioned a hypothetical future job.

Fast forward a few decades: the Silicon Valley sages are preaching that all kids should learn how to program. I noticed in this euphoria an employability undertone. People thought that, to get a good job in the future, children needed to learn how to program a computer. Initially I thought it was a good idea that more kids would develop this skill. But I figured that, because the initiative was misguided, many kids would eventually lose interest. Worst: they would decide they are not good at it. And only the best are good enough for the industry. Because of this pressure, I reasoned, some would start fearing the subject.

In the meantime everyone was eagerly learning how to use some newly developed digital technologies: search engines, email and social networks. Gmail took over vast swaths of the personal messaging exchanges while Facebook became the de facto social arena. Together, a handful of behemoths started to mediate all our communications.

Fast forward yet a couple of decades; the world is taken by a generative AI storm. Suddenly everyone around me is rushing to learn how to use these tools. But we haven’t yet mastered the underlying tools, how can we hope to exercise agency in an AI powered society? How can we expect to become more than milking cows for this world-wide machine if we were unable to develop agency in the preceding, less complex domain of networked communication technologies?

Agency

We all learn how to read and write in school, but only a handful of us become poets, book authors or journalists. That doesn’t mean we wasted our time learning those skills. We all know that reading and writing are not only tools for our jobs. They are, crucially, tools for our civility, for our participation in the democratic process, for organizing ourselves, and much more.

But the recent antidemocratic developments across the world indicate that they are not enough. They are also a dramatic display of what happens when we focus our efforts in uncritically learning how to use the tools that are provided to us. These platforms prey on our most primitive instincts to turn us into clicking and buying machines. To the detriment of everything else.

Unfortunately, my everyday experience shows that computer programming is still only seen as a skill that is useful (if very useful) for the job markets, even though there doesn’t seem to be a difference between the role of traditional literacy and computer literacy anymore (here defined as the ability to write, rather than use, software). Some are probably even thinking that, with machine generated code, no one needs to know how to program anymore.

But programming is not solely a tool for producing software. It is urgent that we understand it as a tool to participate, in equal terms, in our world. If machines are now writing code, it is even more important that we understand it. We can no longer admit reluctance to use version control systems or fear of the command line, much less of programming languages. There will be a time when machines write code we can no longer understand, but, hopefully, at that point we will be the masters of these machines, not passive users.

All of this is especially true for children. I can understand that in a now remote past (or, sadly, some "remote" lands) one generation was literate while their parents weren’t. If you think it’s too late for you to learn how to program (it’s not!) at least don’t deny your children that right. Don’t deny them their agency. AI tools will become easier and easier to learn; programming will continue to be difficult.

I grew up believing I was bad at math. No one told me there is a huge spectrum between not knowing how to count and being a math wizard. One doesn’t need to excel at something in order to have a grasp of it. We don’t need to be a book author in order to participate in the democratic process. But we do need the ability to read and write. And also, I’m arguing, the ability to read and write computer code.

Thanks to the way my parents introduced me to the art of learning I eventually taught myself how to program, out of pure curiosity. Now it pays my bills, but it is also a lot more than just a profession. This is why I’m not afraid of machines taking my job. Real intelligence is indistinguishable from agency.

Looking forward

I don’t believe anyone will be better off in the future by learning how to use big-tech AI tools any more than by using their search and email or social networks. These services are not developed for us. They are built by a handful of people from a narrow demographic group. All of their prejudices and biased are built into these programs. Not to mention their ultimate goal: we can’t expect these companies to respect civic values over profit.

The only way forward is to deploy, from the bottom-up, our own AI tools. Every household, building, neighborhood, school, university, municipality, county and country should have their own server running AI systems, built with free and open source code, available for everyone to read, understand and modify. These programs should employ models trained by their communities of users, with their needs in mind.

But we can’t get there if every single one of us is not involved. Even worse if we’re wasting our time learning how to use someone else’s tools, while, at the same time, teaching and improving them. And we need everyone to speak the same language. The machine language. We should be ever more suspicious of machines that speak our language.

Let’s all learn how to program, but for the right reasons now. We cannot afford to fall for the big tech trap again.